This is a study that will consider the effect that inaccurate statement of age on the death certificate and in the census record has on the published death rates for the very elderly, and in turn, on the life table values for this population. Particular emphasis will be given to examining differentials in accuracy by race and sex. By using a procedure known as the "method of extinct generations", a method that permits the reconstruction of "extinct" population cohorts from the death statistics for a series of years, new estimates of the United States population 85 years of age and over will be prepared for the years 1951 through 1965 for single years of age by sex and race. The resulting estimates for the periods around the 1950, 1969 and 1970 censuses will be compared with decennial census statistics. A series of age-specific death rates will be computed from the derived populations and these will be compared with official data for the United States. The new series of death rates will be utilized in the preparation of life tables for the elderly, and the values obtained will be compared with the official current United States life tables. Trends in mortality and survival among the extreme aged population will be analyzed.